autonomous vehicles: past, present, future · obstacle detection and avoidance ... –urban...

AUTONOMOUS VEHICLES: PAST, PRESENT, FUTURE

CEM U. SARAYDAR Director, Electrical and Controls Systems Research Lab

GM Global Research & Development

GENERAL MOTORS FUTURAMA – 1939 “Highways & Horizons” showed an

imagined world of 1960, complete

with automated highways

AUTONOMOUS HIGHWAY OF THE FUTURE – 1950s

EARLY GM AUTOMATED VEHICLES

Demonstration August 1997 along I-15 near San Diego

Focus on platooning for safety and increased traffic density

NATIONAL AUTOMATED HIGHWAY SAFETY CONSORTIUM

Demonstrated Technologies

Vision-based road following

Lane departure warning

Magnetic nail following

Radar reflective strip following

Radar-based headway maintenance

Lidar-based headway maintenance

Partial automation and evolutionary systems

Close vehicle following (platooning)

Cooperative maneuvering

Obstacle detection and avoidance

Mixed automated and manual driving

Mixed automated cars and buses

Semi-automated maintenance operations

60 miles, <6 hrs, <30 mph

Urban traffic; mixed (human + robot operated) traffic

89 35 11 6 1

GM/Carnegie Mellon “BOSS” finished 1st in 4 hours 10 minutes!

DARPA URBAN CHALLENGE 2007

Low-speed city car equipped with active safety and automated driving technology

Ideal for short distance or “last-mile” personal transportation in inner cities, business campuses, retirement communities, etc.

Outfitted with cameras, GPS, Lidar, maps, V2X communications, smartphone, and RFID technologies

Capabilities

– State-of-the-art autonomous chauffeur

– Autonomous valet parking and retrieval

– Urban automated platooning/traffic jam assist

– Intersection collision assist

– Pedestrian crash avoidance

Demonstrated at ITS World Congress, Detroit MI, September 2014

SECOND-GENERATION ELECTRIC NETWORKED VEHICLE

CADILLAC TO INTRODUCE SUPER CRUISE ON ALL-NEW CT6

CADILLAC SUPER CRUISE

= + Sensor

Fusion

System

3 Short-

Range

Radars

1 Long-

Range

Radar

1 Front

Camera

8-10

Ultrasonic

Sensors

1 Rear Camera

2 Short-

Range

Radars

HOW IT WORKS

LANE FOLLOWING: Using a combination of

GPS and optical cameras, Super Cruise watches

the road ahead and adjusts steering to keep the

car

in the middle of its lane.

COLLISION AVOIDANCE: A long-distance radar

system detects vehicles more than 300 ft. ahead.

The vehicle will automatically accelerate or apply

the brakes to maintain a preset following distance.

ACTIVE SAFETY

AUTOMATED STEERING & LANE FOLLOWING

WHY DO WE CARE ABOUT AUTONOMOUS?

Child Safety

Stay Safe and Secure

Avoid Danger Reach Destination on Time

Door-to-door Transportation

Be Productive Communicate with Others

SOCIETY OF AUTOMOTIVE ENGINEERS – LEVELS OF AUTOMATED DRIVING

© Copyright 2014 SAE

THE AUTOMATED DRIVING PUZZLE

Perception

Legal/ Regulatory/ Insurance

Testing (Verification & Validation)

Human Factors

Social/ Political Economic

Localization

Planning & Decision

Making

Vehicle Motion Control

AUTOMATED DRIVING SYSTEM – CONTEXT DIAGRAM

AUTOMATED DRIVING SYSTEM

Wireless Services

Service Passengers

Driver/Operator

Roadway and

Static Objects

Law Enforcement/ First Responders

Other Road Users

Base Vehicle

TYPICAL AUTONOMOUS VEHICLE SYSTEM

Vehicle Control

Mapping & Localization

Sensors & Signal Sources

Environment Perception

Planning & State

Management

SENSING & PERCEPTION PLANNING & CONTROL

HARDWARE

Sensors (Camera, Radar, Lidar)

Processors (CPU, GPU, FPGA)

Actuators (Brakes, Steering, Gear Select)

Transceivers (Connectivity)

LOGIC, SOFTWARE AND

DATA

Image Processing, Sensory Fusion, Perception, Planning, Behavior

High-definition Maps Real Time Road Conditions

SYSTEMS INTEGRATION

Validation and Testing

MANUFACTURING

Assembly and Programming

AUTOMATED DRIVING TECHNOLOGY ELEMENTS

Geographic Location and Road Geometry

Freeways

City center local roads

Arterial roads

Residential local roads

Industrial local roads

Parking lots/parking garages/ residential driveways and garages

Tunnels, covered/multi-level bridges

Construction zones

DIMENSIONS OF STATE-OF-THE-ART: AUTOMATED DRIVING OPERATIONAL DOMAINS

Environmental Conditions

Road surface conditions – Clear

– Wet/puddles

– Snow covered

– Icy

– Pot holes

Illumination – Daylight

– Dawn/Dusk

– Night

Atmospheric conditions – Clear

– Rain

– Snow

– Fog

– Blowing dust/leaves/debris

DIMENSIONS OF STATE-OF-THE-ART: AUTOMATED DRIVING OPERATIONAL DOMAINS

Traffic Conditions

Density – Light (flowing at or above speed limit)

– Moderate (continuous flow, below speed limit)

– Congested (surging, stop and go)

Speeds

– Very low speeds (e.g., up to 5 mph)

– Stop and Go (e.g., up to 20 mph)

– Local road speeds (e.g., up to 35 mph)

– Arterial road speeds (e.g., up to 50 mph)

– Freeways (e.g., 80 mph)

DIMENSIONS OF STATE-OF-THE-ART: AUTOMATED DRIVING OPERATIONAL DOMAINS

Parking

Driver/operator role and position – In vehicle, beside vehicle, or remote

– Supervised or unsupervised (valet)

Parking environments – Street side

– Parking lot

– Parking garage

– Residential garage or carport

Parking spot – Parallel

– Angle

– Back in

DIMENSIONS OF STATE-OF-THE-ART: AUTOMATED DRIVING OPERATIONAL DOMAINS

Production-viable sensing/perception

Fault-tolerant/fail-safe automated vehicle control (with driver-in-the-loop)

Situational analysis in complex environment

Emergency situations and rare events

Dealing with diverse behaviors of others (non-autonomous vehicles)

Detection of driver distraction (inattentiveness)

Positioning – Digital maps with lane-level accuracy, road signs, etc.

– GPS lane level accuracy and availability (urban canyons,…)

– Localization with limited accuracy or no GPS

Virtualization – Physics-based active sensor models

– Verifiable driver (non-robot) model

V2X: Security/privacy, interoperability, congestion

AUTONOMOUS DRIVING CHALLENGES AND OPPORTUNITIES

Sensors

RADAR

– Long Range – 120m x 14º

– Medium Range – 70m x 90º

– Short Range – 30 m x 150º

Video

– Mono and Stereo

– Visible and IR

– Front and rear

LIDAR

GPS/map databases for navigation systems

Actuators Controlled by Computers

Electric Power Steering

Brake Systems(Antilock Brakes/ Traction Control/Stability Control)

Engine and Transmission

Communication Networks

CAN, Flexray (for safety critical systems)

Ethernet (for infotainment)

ENABLING TECHNOLOGIES IN CURRENT PRODUCTION VEHICLES

21

Sensors

Object Sensing – Smaller/easier to fit on the vehicle

– Less expensive

– Higher resolution (range, horizontal and vertical angle)

– Larger field of view (longer, wider)

– Higher update rates and lower latency

Road Sensing – Sign/traffic signal information

– Lane geometry

– Surface friction

Driver State Sensing – Attention

– Intent

Fail Operational Functionality

Sensing

Actuation

Processing

Communications

Power

Networking and Infrastructure Information

Maps/GPS – Lane level information

– Faster update rates

Vehicle to Vehicle and Vehicle to Infrastructure Communication

ENABLING TECHNOLOGY NEEDS

THANK YOU!

Cem Saraydar

cem.saraydar@gm.com